Bitcoin Forum

The events of the latest software bug which catalyzed a massive 23% flashcrash, show us that bitcoin is not a protocol yet - it's still totally bootstrapped to the open source Satoshi client. One bug affects the entire network. It's not a protocol, it's whatever that piece of software determines it is.

This doesn't happen in any other p2p network that I know of. If there's a bug in some filesharing software, the whole network doesn't go into a panic. Only the users of *that* software have to deal with *their* problem.

Of course, I believe there shouldn't have been an orchestrated revert to the "old" chain (a.k.a. an orchestrated 51% attack). I believe the market should have solved this on their own. Yes, it would have been a massive blow to the miners who didn't upgrade, but it's their fault for using a client that follows a different protocol. That would have reaffirmed that the software you use is your vote for which protocol to use.

You see, even the protocol isn't set in stone. If the majority of hashing power (or whatever non-full nodes) vote for a new protocol and use software that implements it, then that's the new bitcoin. That's what makes bitcoin _decentralized_.

Instead, we have one client, one source, one community. This is not decentralization.

This is what ultimately makes bitcoin vulnerable to human imperfections. The press is going crazy with "bitcoin is not stable thanks to a software bug". How can we possibly expect people to trust the entire monetary system on one piece of software written by human beings?

I will point out that we're still in a very immature stage, and I believe these stuff will sort themselves out. But I just wanted to clarify how early a stage we're still in.

I'm mostly seeing commentary on how quickly bitcoin recovered.

The majority of hashing power voted, and the vote was: rollback and revert to 0.7

Like I said, the crash itself doesn't worry me. It's the immaturity of the protocol which currently has virtually only one client. This immaturity means two things: bitcoin does not yet fulfill it's promise as a decentralized currency and it's not ready for prime-time where software bugs can have such huge repercussions in the market.


Fair point. But notice how easy it was to pull off this 51% attack. Without Gavin's decision making, the new chain would have taken and a totally different course of events would have taken place.

By flash crash, do you mean a crash that is over in a flash? As exchange rates have been looking pretty normal to me, other than a bit of a dip that was over in a flash.

Oh we didn't get back to $49.xx yet? That is normal too, touch a nice high, someone gets cold feet, step down a little to prepare for a more leisurely assault of the heights, I don't see a problem. ???

-MarkM-

Wow, do you even understand what happened?

The market *DID* solve this on its own. Calling what happened an "orchestrated 51% attack" is an awfully dramatic manipulative misinterpretation of events fit mainly for irrational "sky-is-falling" tinfoil-hatters.

Eleuthria switching BTC Guild back to v0.7 was the hash-power that broke the camels back, as it were. The chains were both at 40+%, and Eleuthria had enough hashing power to get v0.7 back over 50% if he switched from v0.8 to v0.7. That is NOT a 51% attack. Stop calling it that; it's just wrong.

OMG! Read the f*cking IRC logs. Gavin's instinct was to NOT switch back to v0.7, but other people thought going with v0.7 was better, Gavin eventually agreed (as one of many relevant voices, and certainly not dominant), and the folks with the best input on the subject all gave the "ACK" for Eleuthria to go ahead and revert BTC Guild, which as I pointed out above, merely provided the marginal hashing power to get v0.7 over 50%.

<tinfoil hat>

OMG it was a 10%-or-thereabouts attack! We always knew pools were a bad idea!

Just one pool can launch a 10%-or-so fork-merging attack! Somebody do something!

</tinfoil hat>

Please stop the "Gavin's decision" meme, too:  I went with the in-the-moment consensus when it became clear that it was POSSIBLE to switch to the 0.7 fork.

And as Melbustus said:  that was only possible because the split was close to 50/50.  If more miners had already upgraded to 0.8, an alert would have been sent to non-0.8 peers telling them to either upgrade or shutdown until we could find a workaround for the problem.

It is called having an open mind as well.  Listening, analyzing and responding as new information is available instead of digging in ones heels.

I was very proud of how well all of the developers handled the situation and how even people with different opinions on some issues were able to put them aside for the best interest of the community. 

This event actually INCREASED my confidence in both Bitcoin and the Bitcoin community. 

I agree that it is a big problem with bitcoin when one client dominate the market. If new client is not pooping up soon we will see start to end of bitcoin. If we hade 100 different clients than it wasn't easy to make a decision in a IRC channel with small team involved and with few people more active than other (It is not market , It is Centralized system or core team model).

+1. Well said.

OP: you're right. But I think you miss the point of the Bitcoin project: to create a decentralized payment processing network. Your observation would be better put: "we're not done yet!"

I just want to emphasize, I have no criticism against anyone in the bitcoin community. I most certainly did not mean to make the particular point that I would call what happened as technically a 51% attack (and it wasn't a bad thing, it was simply the decision that was made at the time and bitcoin is still alive and kicking). Also, I did not mean to make Gavin look like a terrible dictator, only that I see it as there being a focal point where network-wide decisions are made (and this is not a bad thing, it's simply the way it is and bitcoin is still alive and kicking). Like I said, everything is fine and I believe bitcoin was designed in such a way that things will eventually all work out (until a competing currency takes it's place).

My point with this thread was to simply emphasize that the network is still young, as evident by the fact that this decentralized network still relies on a very strong (and excellently performing) developer community. I think they're doing a great job, but please note that as long as we need to rely on the developers of the single most dominant client, the network is still very immature.

The protocol is not yet a protocol if there's only one de-facto client using it. In this case, it's simply the client's implementation, not a protocol.


Notice the title has a "yet" in it  ;)

Wall Street Flash Crash

Thursday May 6, 2010 in which the Dow Jones Industrial Average plunged about 1000 points (about 9%) only to recover those losses before the end of the trading day. It was the second largest point swing, 1,010.14 points, and the biggest one-day point decline, 998.5 points, on an intraday basis in Dow Jones Industrial Average history.

Einstein, what do you think happened?  ::)

Or did you see anyone holding a gun to the heads of miners forcing them to follow their orders?


You are misinformed.

Had many miners not switched back to 0.7 no one would have been forced to do anything, we would have just had two hard forks of the blockchain and those who favored the 0.7 fork could use that and those who favored the 0.8 fork could use that. Of course who knows what that would do to BitcoinA and BitcoinB and their usefulness but my point is that the quoted above is factually incorrect.

There is another threat suggesting the documentation and "formalisation" of the protocol, some of the devs are reluctant as it means lots of work lots and lots of documentation for something that will probably change dramatically over time and the fact that whoever writes the protocol is doing so as a small elite so to speak.

I think if it was atleast documented and not quite standardised but suggested then we could have an actual protocol instead of this situation where the original client code is the protocol.

If there was this documentation available it would attract more devs and would stop threads like this springing up, Clients would be able to be developed independently that are compatable with each other and bitcoin would prosper, if its documented people can review and comment on it even if they don't understand C++.

So, like any government in the world, now we have the left wing and right wing, bitcoin users finally have a choice  :D

I'm still not very clear how this fork will affect the merchant, will their business transaction get random error or they can select which chain to use by simply switching to a different client version?

That's what happened.

It might mean lots of work for someone, but what devs are we speaking of here, documentation devs or code devs? I suspect if the doc devs just shut up and got to work on the docs the code devs wouldn't mind at all.

And if the code devs feel the docs devs are a small elite and that there is something wrong with that they are welcome to wear multiple hats, aren't they? Or are the docs devs opposed to permitting code devs to participate in their hallowed process?


I am interested in seeing some alternative code in action, too. Maybe the docs devs can document what protocol-critical (hence "part of the spec") behavior in the various other impementations out there differs from the Satoshi-node behavior so folks can weigh which behavior best fits "the spec" or would make a better "spec" ?

-MarkM-

Bingo. I agree.  This experiment is still in progress.  A true protocol it is not.  Is that the goal? Yes.  But we need to survive until we get there.  The 1.0 release will be very interesting...hopefully it won't come too soon.  We can't have oopsy network splits in 1.0. 

Just a decentralized payment processing network? I think bitcoin is much more than that  ;)

As an end user, I don't care too much about the transaction system being centralized or not, as long as it provide certain function. Bitcoin can not protect consumer (charge back not possible), it can not replace other payment services like VISA/PAYPAL

Many people are attracted by bitcoin because of the fundamental flaw in today's debt-driven monetary system since the abandon of gold standard  (unsustainable growth) , there is a strong demand for a type of honest money

All those claiming that the market sorted itself out, you have a valid point and I mostly agree. The market did sort itself out.

I just have one question: what would have happened if there had been no time-sensitive discussions among developers? What if the new version were released and the developers would not have had discussions as to what should miner do or not do?

I'm genuinely curious if the miners would have reverted to the old chain. Again, I don't think anything "wrong" happened, only that I think the miners of 0.8 should have stayed on 0.8, making other miners understand that it's not the software that determines what's good or bad, it's simply that your software has to speak the same language that the majority of software speaks. Hence, a protocol. Perhaps they thought that moving forward with the new chain would be detrimental to the future of bitcoin and the future of their mining enterprise. I'm just curious.



"The protocol" is the language which most software understands. Nobody "writes" the protocol, just as nobody "writes" the English language. If everyone starts using this new word like "tweeting" then that's now part of the English language. Even if no "small elite" determined it. Like you said, the original client is the protocol - but not because that they are writing it, simply because they are the only ones speaking this language. To decentralize you simply need more people speaking the language.

What you're asking does not make any sense because to me you are asking what would the market do if the market couldn't act in the first place since obviously when it could act, your scenario didn't play out.

If the market can't act then the market can't act and Bitcoin, in my eyes at least, has already failed regardless what that would means specifically..

Maybe when the market can't act, it sings and dances! :)

(Lots of song and dance from people who [ didn't | don't | weren't equipped to | aren't equipped to ] act ...)

-MarkM-

More documentation is great, so yeah, if you want a formal spec, go for it.

Here's a tricky question you can start with:

Assume there is a fork consisting of max-block-size blocks. How deep a fork/re-organization MUST a conforming implementation handle?  6 blocks? 1000 blocks?  as-many-blocks-as-there-are-in-the-chain blocks?

Does that imply that a confirming implementation MUST be running with a certain amount of memory, or MUST a conforming implementation be able to handle such a chain fork within a certain amount of time?

.. and once you answer all that:  what if the network consists entirely of non-conforming implementations that take shortcuts and just assume that there will never be a re-org more than X blocks deep?

That part is probably the easiest part. It is merely a description of an archaeological dark age that preceded the advent of the doc devs' excellent and so very much needed work.

Looking forward to seeing what they come up with! :)

-MarkM-

EDIT: a very clunky "shutdown if ever a re-org of more than 6 blocks would otherwise be attempted" might date back to an era in which Pak Protectors (http://en.wikipedia.org/wiki/Pak_Protector) were still around expecting to intervene to fix anything at any time, as is their wont.

Unfortunately there is a lot of FUD floating around about the supposed "risks" of alternative fully-validating implementations.  If you take a calm, careful look at the technical issues involved and don't settle for slogans it turns out that using an alternative fully-validating implementation carries no new risks (https://bitcointalk.org/index.php?topic=134318.msg1441171#msg1441171) (relative to the Satoshi client) except possibly the loss of revenue by people who mine with said alternative implementations.

This assumes that the alternative implementation takes a few sensible precautions: (a) watch for long invalid chains and if you see one go into safe mode and (b) watch for X-block-long forks within the last day or so and if you see one go into safe mode, where X is the number of confirmations you wait for before believing a transaction and a "day or so" is longer than it takes the miners to notice that they could be losing money (empirically: less than two hours).

The Satoshi client already does (a) and always has.  If it had done (b) the recent blockchain fork would not have happened (this is hindsight -- I'm not claiming to have noticed this beforehand).

As long as they go into safe mode on detection of a long invalid chain (like the Satoshi client does on this line of code (https://github.com/bitcoin/bitcoin/blob/1a9ee5da327d8079a297ad292a1c16745b75df91/src/main.cpp#L2944)), they'll simply get stuck until a human intervenes.  Not pretty, but it could be worse.  In the meantime the miners will fight it out, like they did during the recent fork, with a powerful incentive to come to some sort of agreement since a lot of hashpower (money) is being wasted in the meantime.  If there's a dispute, eventually somebody will cry "uncle" and relent.

Regardless, on the larger point I agree that trying to call some prose document a "specification" is not really practical.  The answer always will be "whatever most of the clients will go along with."  Longer-invalid-chain and recent-long-fork protections are important because they make this more predictable.

Related : http://bitcoin.org/may15.html

You are wrong about one client. Ufasoft Coin also is full validating node and miner too.

Great, so we have at least two implementations.

Anything network-critical they do differently presumably implies one or both of them are not up to spec. What they both do might happen to match the spec.

So the doc devs should be able to get a first draught of a spec down just by documenting what it is that both of these do.

-MarkM-

We would have reverted, because Eleuthria started the roll-back process on BTC Guild prior to receiving an ACK from the devs (by at least a few seconds). The only way this would have turned out differently would have been if the devs were insistent on not going back to the pre-0.8 chain and if Eleuthria listened to them.